O. N. Granichin, V. I. Vasiljev

Initial model for the description of classical algorithms is well-known the Turing machine. The realizability of algorithm for such machine appears is equivalent to final number of steps (the moments of movement of a head of the automatic device). R.Fejnman has proved impossibility of representation of results of quantum mechanics on such computer even with use of likelihood algorithms. It is connected by that the number of elementary operations will grow экспоненциально with growth of number of degrees of freedom of system. In computer science within the limits of the theory of complexity of algorithms a lot of the results proving theoretical impossibility of the effective decision of many significant problems from the practical point of view, in particular, of natural processes connected with modelling is received. It means, that at attempt of the description of process interesting us with the set accuracy by means of the Turing machine its alphabets of conditions and memories and-or time of functioning appear developers of systems of automatic control who conduct are inadmissible are great. One of the first with these theoretical difficulties have collided researches on a joint of computer science and this or that concrete scientific discipline (physicists, biology, chemistry, etc.). At the same time realizations of a lot of modern methods of the theory of automatic control do not keep within frameworks of discrete model. One of possible ways of its generalization is use so-called «hybrid systems «which properties are actively studied recently. In practice of automatic control there was fruitful the idea of relay regulation consisting switching of a control system from one «sliding» mode on another. Into last decade in control systems actively takes root neural the approach. Often speak about use in operating contours neurocomputers. But work of the listed devices is not described completely by the classical scheme of Turing machine. One of variants of expansion of thesis Church-Turing is introduction of new concept of a computer in which calculations are under construction not on traditional algorithms, and on models. A following natural step - generalization on model of dynamic systems. The first, the basic and the most disputable, from the modern point of view, concept of the discrete scheme of the Turing machine is the concept «step». Another importance restriction of classical model of calculations is splitting memory into the isolated bats therefore as, first, reduction of length of a step and distances between bats from the certain level does impossible to consider them separately by virtue of laws of quantum mechanics. Refusal from «scalar bits «will allow to speak, maybe, about performance of multivariate (vector) operations for one «step». For example, Item Shor has offered algorithm of quantum transformation Furie, which can be carried out in time proportional , instead of for as classical fast transformation Furie. Authors had been offered and proved the new abstraction of the computer including all listed above approaches. The new concept essentially differs from classical inclusion «continuity» evolutions and reconsideration of concepts «tape» and «a cell of memory «. If in the classical approach of a cell of memory are used for storage of the discrete information and their change is possible only when «tape» in the memory having the heterogeneous nature to which the general can be applied to all cells the program of evolution shows the index of a tape on it in the new concept «the cell of memory «represents constantly functioning model of any dynamic system (probably, enough complex), and. It is obvious, that the classical cell of memory for storage «bat» is a special case of such generalization. In any modern computers storage of the information is based on those or other physical principles, only evolution of a condition of a cell during storage more simple - is kept by a constant any physical characteristic. In other words, the traditional computer science can be compared to arithmetics, it operates with figures - values of cells of memory (or, it is more exact - arithmetics of final binary fractions). The offered model of calculations will allow to pass in computer science from «arithmetics» to «the functional analysis «, investigating processes of evolution of the information inside of new «cells». Oписывается mathematical model of new processes of calculations as sets of working in parallel dynamic systems with crossed parts of phase spaces (the general memory) and «spasmodic» switchings between them. In the third section the formalized description of generalized the Turing machine including specification of earlier attempts of the description of similar system is given. Further universality of new model is shown. The offered new model of calculations allows to describe if not everything the overwhelming majority of processes of the real world, and also work of every possible existing and future computers, including analog and biocomputers, neurocomputers, quantum computers, etc. Feature of the offered approach is refusal of a reduction of complexity during calculation. Thus, concept of computing complexity to consider concerning the chosen system basic evolutionary primitive more correctly, instead of it is rather traditional considered bit transformations {0,1}. Quantum and neurocomputers promise to change strongly representations about computing capacity of modern computers. The increase in computing capacity, possible due to use of new models of the calculations which are based the physical phenomena, allows to assume, that in the future new computers can solve problems, impracticable for usual computers.